Fault Monitoring and Location Strategy for Power Communication Network Based on Multidimensional Data Association Fusion

To make full use of the massive data contained in the power communication network for fault assessment, a fault monitoring and location strategy for power communication network based on multidimensional data association fusion is proposed, which includes processes such as multi-source data fusion, massive data filtering, and clustering fault research. A massive data filtering method including two steps of time-aligned filtering and feature value hit filtering is proposed to filter out a large amount of irrelevant data. A dichotomy-based k-means initial cluster center selection algorithm is designed to improve the cohesiveness of clustering clusters. This strategy can analyze the specific network devices and links where failures occur from a large number of complex and diverse network alarms, and accurately assist in fault recovery.